[go: up one dir, main page]

Service status

Check
curl -s https://api.factlot.com/health
Answer
{"ok":true,"version":"0.5.0"}

How to check whether FactLot's hosted service is up, what is watched, what has gone wrong before, and how to tell us about a problem. This page shows no uptime percentage, because FactLot does not measure one yet.

Trust pages: Status · Security · Service levels · Assurance. Written 7 November 2026.

Check it now

This page runs no scripts, so it cannot show live status. The live check is the service's own health endpoint:

curl -s https://api.factlot.com/health
{"ok":true,"version":"0.5.0"}

api.factlot.com/health answers with "ok": true and the running version when the hosted API is up. It needs no key, and it counts toward the free tier's 60 requests a minute like any other request.

The version above is the one this page was built from. The number changes only when a release is tagged, so the hosted service can run changes that the changelog lists under "Unreleased" before the next tag.

What is checked, and how

Health endpoint

GET /health, answered by the API itself. It shows that the API is up and which version is running. It does not test the registry or billing. How often: whenever anyone calls it. Nothing polls it for us.

Live suite

npm run test:hosted (scripts/e2e/hosted.ts) replays every shared test vector against the hosted REST API, calls every MCP tool and checks the MCP answer equals the REST answer, does real registry lookups in the hosted regions, and fetches the main site pages and the home page's links. It exits with an error on any failure. The last full run recorded here, on 8 November 2026 after the hosted registry opened, passed 4,635 checks and failed none. Shipped after that run: paid gating on the free tier, the optional free key, the usage counters and the nine-region registry release and the registry releases after it. How often: run by the maintainer after a deploy. It is not on a schedule, so it shows the state at the time of the run, not today.

Continuous integration

The CI workflow on GitHub Actions runs lint, formatting, type checks, the TypeScript and Python tests, the site build, the MCP end-to-end test, OpenAPI validation and the TypeScript-versus-Python differential test (details). It checks the code, not the running service. It is linked as text, not as a status badge, because a badge image would make your browser contact GitHub when you read this page. How often: on every push to main and every pull request.

Worker logs and errors

Cloudflare Workers observability is on for the API. A failed request logs only the error's name, never its message or any request content. Usage records hold the time, key ID, plan, path, method and status, with no IP address (see the privacy notice). How often: continuous; a person reads them when something is reported.

There is no automated uptime monitor in the repository, and nothing in the repository sends an alert when the service is down. An outage that lasted minutes and that nobody reported would not appear in the history below. That is why there is no availability figure on this page, and why the service levels do not promise one yet.

What FactLot depends on

The hosted service runs on Cloudflare, bills through Stripe, and keeps its source on GitHub. When one of them has an incident, FactLot can be affected, and their own status pages are the first place to look:

Incident history

Incidents since the hosted service opened
Date What happened Who was affected Cause and fix
None recorded since the hosted service opened on 3 November 2026.

"Recorded" has a limit, explained above: without a monitor, only an incident that someone saw or reported can be recorded. When one happens, it is added here in plain language with the date, how long it lasted, what was affected, the cause and what changed. Bugs that were found and fixed without a service interruption are not incidents; they are in the changelog under "Fixed".

Report a problem

  • Service down, slow or wrong: email contact@factlot.com with "STATUS" at the start of the subject. Say when it happened (with time zone), the URL or tool you called, the HTTP status or error you saw, and your key ID if you have a paid plan. Never send the key itself.
  • A wrong number from a calculator: see how to report a wrong number.
  • A security problem: do not use a public channel; follow the disclosure process.
  • Anything that is fine to be public: GitHub issues or discussions.

Response targets for paid plans are in the draft service levels.